Flight Plan: Steel Manning
Counterintuitive but true: The best replies begin by strengthening your opponent's argument first
This is the second installment in the Flight Plan series, which seeks to bring the foremost reasoning strategies into the age of AI, adapting them for use with a powerful, but sometimes confused co-pilot.
One of the best ways to improve your reasoning is to elevate an argument before responding to it.
I am always impressed when I see someone do this consistently. Choosing effort over the internal inertia that favours shortcuts demonstrates a genuine commitment to getting things right.
The Principle of Charity is one expression of this practice. Most famously advocated by the philosopher Donald Davidson, it holds that communicating effectively depends on interpreting what others say charitably. That means assuming others, even those with wildly different views than your own, are rational, logical and by-and-large, believe mostly true things.
Being charitable is hard. It means that no matter how implausible an argument sounds, you shouldn’t reject it outright. You should consider what reasons, logic, and beliefs might lead to that conclusion, and engage with those in your response.
Even better is steel manning. This is the practice, advocated by philosopher Daniel Dennett in his book Intuition Pumps and Other Tools for Thinking, of not only assuming best intentions on the part of your interlocutor, but actively improving their argument. It is the opposite of a straw man, an argument deliberately designed to be weak so that it can be knocked down. The steel man is the strengthened argument: if you can knock it down, your opponent has nowhere left to hide.
Steel manning takes discipline and energy. You must dip into the psychology of another, adopt their point-of-view, and live inside their argument long enough to improve it — and you must do this knowing that your final act will be to crush it.
That’s what makes it so impressive … and so rare.
Here is a skill where I think using generative AI makes sense. It is not only better at steel manning, but because it does not fatigue, can do things to improve an argument that would drive the average person around the bend.
One of the key advantages of using generative AI to steel man is that it can easily take on the persona of others. It is ever mutable in its beliefs and point-of-view. This is where it is better than you. A chatbot can, as Dennett recommends, help you to rewrite your opponent’s view so well that they may say, “I wish I’d said it that way!”
But more than simply rewriting the argument, chatbots can go further. LLMs can map the broader terrain of related arguments, not only steel manning the argument presented, but offering up an alternative that may be better. This extra step — seeing what else is out there — makes AI an invaluable partner when developing your own response. Depending on the context, it may be wiser to replace the original argument entirely, responding to the best out there, not just the one initially offered.
And the partnership does not have to end there. Generative AI can continue to help once you have decided how you will respond. It can steel man whatever counter argument you produce — fortifying your own response as well.
In time, you may find that the most rewarding arguments aren’t the ones you win easily, but the ones where you take the other side seriously.
Steel Manning
What is it?
Steel manning is the practice of strengthening another’s argument. It means making your opponent’s case as compelling as possible, sometimes even better than they originally stated it. In short, it’s the discipline of elevating an argument before offering a response of your own.
Why does it work?
Steel manning works because it limits how an opponent can respond. By knocking down their strongest case, the other side is left to choose from a menu of weaker options. They can try to mount a new attack, but those are easier to beat back once you have trumped their highest card.
It also has a multitude of benefits for you, the arguer. By digging deeper into what someone else believes, you gain a clearer understanding of your own argument. Your own assumptions, strengths, and weaknesses are reflected back at you, offering new insight into how your premises, evidence and conclusions fit together.
When should I use it?
Whenever possible. Getting into the practice of cleaning up the thinking of others makes you a better thinker.
It matters the most when the stakes are high. If you must win an argument then steel manning ensures that you don’t waste time responding to something that is easily brushed aside.
How do I steel man?
The following prompt has been developed with the help of generative AI to ensure that it is sufficiently detailed and follows best practice.
Before running the prompt, review the pre-flight checklist.
Clarify the argument: Consider asking a chatbot to put your opponent’s argument in standard form — a list of premises that support the conclusion — to check whether you have the same understanding of the argument as the bot.
Set boundaries: Decide how far the model should go in improving the argument. Should it stick closely to the original point of view, or is it free to explore stronger alternatives—even if they differ significantly from the original framing?
Focus the task: Are you looking for a complete upgrade to the argument, or just for it to identify weak spots or missing evidence?
Do not hesitate to alter the prompt below to ensure that the output matches your needs.
Flight plan: Steel manning prompt
Act as a skilled advocate. Your role is to strengthen the best possible version of the argument provided—regardless of whether you agree with it. Your goal is to make the argument clearer, harder to refute, and more persuasive, as if it were being made by someone deeply committed to its core idea.
Your mindset: Approach this task with intellectual honesty, clarity, and generosity. You aim to present the argument so effectively that even its original author might say, "I wish I had said it that way."
Instructions – Structure your response with the following clearly labeled sections:
1. Improved Argument
Present the strongest version of the original argument. Use a short paragraph or standard form (premises → conclusion) for clarity and logical flow.2. Improvements Made
Bullet-point the specific enhancements you applied—such as clarifying terms, filling in missing reasoning, improving structure, or strengthening evidence.3. Limitations or Constraints (if applicable)
If the original argument was vague, confused, or based on flawed assumptions, state this respectfully. Clarify how you interpreted and improved it while staying faithful to its intent.4. Alternative Stronger Argument (optional, but include if applicable)
Reflect on the broader belief system or motivations behind the original argument. Then, if applicable, present an alternative version of the argument that someone with the same core values, worldview, or goals might reasonably adopt—one that is more persuasive, logically coherent, or defensible.Include the following:
Alternative Argument: Present this new version clearly.
How It Differs: Briefly explain how this version reframes or upgrades the original.
Why It’s Stronger: Explain why this alternative better serves the same underlying aims—e.g., avoids common objections, aligns more closely with evidence, or increases persuasive power.
Original Argument:
[Insert argument here or attach it]
Try it!
I think the easiest way to try this prompt is to grab an article defending a position that you disagree with and use that as a starting point. You might be surprised with how you react to a position that you consider wrongheaded when it is presented in the strongest light possible.